According to Wayne Zimmerle’s MSA program, I see that Nicholas has defeated 9 different masters for a total of 13 times. He also has 13 draws against masters, including four new names.
wins: FM Ronald Cusi, Emmanuel Perez (3), Michael Aigner, Albert Rich, FM Daniel Naroditsky, Paul Gallegos, FM Eric Schiller (3), Igor Margulis, Jerry Hanken
draws: FM Ronald Cusi, Emmanuel Perez (2), Richard Koepcke, FM Frank Thornally, Sergey Galant, Paul Gallegos, FM Eric Schiller (3), Igor Margulis and Russell Wong (2)
His lifetime record against masters is +13 =13 -19 for 19.5/45 = 43%. That’s remarkable when you realize that the rating difference exceeded 200 points in 19 of those 45 games, with Nicholas being the lower rated in every game.
Ah, I see what I was doing wrong, I wasn’t making sure I always had the opponent’s regular rating instead of his quick rating. (This was good practice for when MSA is rewritten.)
I reran my program that ‘walks’ the crosstables, I also find his record is 13-19-13 against masters based on their pre-event ratings and 12-19-14 against 2200+ players based on their post-event ratings.
FWIW, his record against experts (pre-event ratings) is 32-32-28.
This topic isn’t very interesting to me since I have such doubts about the rating system.
Take for example a player (someone I don’t know) in the current Foxwoods Open. He’s currently ranked 26th and is about 2125 and has had a pretty good tournament with a PostRating of 2145 (or so).
How did he make that progress and is the change in rating reflecting his abilities? He beat several masters which should show his ability as at least 2300. But, that’s far above 2145 (or so).
He’s gained something like 20 rating points over 7 games – that’s only about 3 or 4 pts. per game. So, to catch up to his true ability (of say 2300) he’d have to gain about 200 pts at 4 per game. That’s 50 tournament games. If he’s anything like me he only plays about 10-15 games per year, so we’re talking about 4-5 years of play with an incorrect rating. If he plays more often, say 30 games per year then he’d have the wrong rating for about one year.
I wonder how his opponents will feel losing rating points to him during that year or 3 where he’s under-rated.
Same goes for Nicholas and other up-and-coming youngsters.
Fix the rating system to allow for quicker changes and then we can talk about whether Nicholas is truly NM strength yet. I’ll bet the 22xx and 23xx players he’s beaten would probably tell you he’s been there for some time already (at least on occasion).
I’ve had a similar experience. I had hit my floor of 2000 after playing in some lackluster tournaments. I was angry enough that I applied myself more for a while, and got up to 2101 in six months or so. Under the old system, I would have been closer to 2170. This has discouraged me, and I’m lackluster again. It just takes longer at my rating to make any improvements.
Joe, since 1/1/2006 you have played in 51 regular rated events. Your rating went up in 29 of them, it went down in 22 of them. Your average loss has been 6.5 points, your average gain has been 9.9 points. There have been no events in which you earned bonus points during those 2 1/4 years.
Your argument appears to be that you would have gained more ratings points under the old stepwise K formula. However, you would have also lost more ratings points in 22 of those events.
By comparison (for Mark H’s sake), Nicholas Nip has played in 62 events since 1/1/2006. He gained ratings points in 44 of them and lost ratings points in 18, including two events so far in 2008. His average loss was 14.2 points, his average gain was 19.9 points. In 9 of those events, he performed well enough to earn bonus points.
Sorry, I just don’t see any evidence that the rating system isn’t performing reasonably for either Joe or Nicholas.
What else has he done, say against experts, A players, etc, as well as against other masters (not just the ones he has defeated)?
I, for one, occasionally beat masters. I also occasionally lose to 1400 players. My rating remains solidly in class A, and I have no delusions about being master strength.
An easy and logical fix for this problem would be to have the threashold for bonus points decrease as the floating K decreases instead of being a fixed number; has the ratings committee consider this?
I don’t know what the committee has considered, but I don’t think your solution is very logical (or at least you haven’t made a logical argument for it yet), and as a result it might not be very easy to decide what sliding bonus scale is appropriate, either.
As I recall, somewhere around 25% of the performances we have rated in the last year or two earned a bonus, and some of those bonuses were in excess of 100 points.
Moreover, when I looked at expected vs actual performance for the RC last spring, expected performance was a very good fit with actual post-event ratings.
MarkH gives a very persuasive illustration of a player with a 2300 performance currently rated at 2145 and how slowly he will progress to his performance rating, so this is an example of an imperfect fit. I'll bet there are a significant number of others, especially of players rated 2000 and above. Can you check ? Further, I believe that the magnitude of the bonus is tiny at the upper rating levels as a percentage of pure rating points gained,can you check this as well? I think the results will justify my proposed sliding scale fix.
Without knowing exactly who he’s talking about, I cannot say whether MarkH’s example is illustrative of anything except that players win games against players rated well above them. They also lose games to players rated well below them.
Every month there are players who have a performance rating 250 points above their current rating. Most of the time they will get bonus points for those performances. Not all of them are 250 points underrated, sometimes they just had a really good day or one or more of their opponents had a bad one.
A number of years ago I defeated the reigning Nebraska champion, a Master. Was I a 2300 player that day rather than somewhere in the 1500’s or 1600’s? Of course not.
A few days before the event in which Nicholas Nip became a Master, he had an event in which he lost to a 2115 player and drew a 2343 player. So what was he that day, a 2300 player (since he drew one) or a 1900 player (since he lost to a 2100 player?) His rating went up one point in that event.
Granted. However, there are players that have several good days/tournaments and the ratings formula for such fast improving higher rated players is very slow to recognize this in contrast to its treatment of lower rated ones. The fixed bonus threashold ( I believe it is about 30 points) coupled with a smaller K makes it extremely difficult for them to get bonus points. I continue to believe the formula that awards bonus points should be linked by some percentage method to their lower K and that a focused study will prove this is a problem.
CoachBob, I would suggest you read the mathematical description of the rating system. The current bonus multiplier is 10 (where it has been since at least 2002), so for a 3 or 4 round event someone starts earning bonus points when his or her rating increases by 20 points, not 30. The Ratings Committee chair has asked for some data to help them in their annual decision whether or not to change the bonus multiplier. (However, the data he’s asked for this year will be more time-consuming to generate than in past years.)
I still don’t think you’ve made your case regarding a more drastic change in the bonus formula, but I don’t think this is the proper venue for such a discussion, I suggest you contact Prof. Mark Glickman.
In 2007 there were 43,581 performances out of 227,802 (19.1%) that received bonus points. The highest player to receive a bonus during 2007 was rated 2635.
Robert Hess is only 16 and beat 4 GMs at Foxwoods. His rating is 2475 (or so) and the GMs were 25xx-26xx. How long 'til his rating catches up to his performances?
How long is his rating going to be below this obvious 2600 (or 2700) USCF performance?
Look at his performances in the last 2 or 3 years and see how many times he had bonus gains and how many times he defeated players say 200 pts. above him versus losing to players 200 pts below him. If he’s winning a significant number of games against those above, but not losing many to those below, then it’s not just inconsistency, it’s improvements which need to be reflected in the rating.
Is his rating ever catching up to his performances?
For players who aren’t playing as often there could easily be inconsistency. For older players who have reached a plateau there could be fewer big gains or losses, but that wouldn’t matter. Their rating would be as stable as their performances. For players who are getting much older and their performances are decreasing, then there are the somewhat awful rating floors.
Let’s focus on the improving young players who have developed established ratings in the apartheid scholastic system and then move into the ‘adult’ tournaments.
As touched on previously in this thread, just about the only way for a master to gain bonus points is for an IM or GM norm level performance.
I see first hand kids who are rapidly improving and whose performances exceed their rating by 250 or more points. Just this past weekend in Reno, Daniel Naroditsky (2266) had a 2600+ performance and Steven Zierk (2099) had a nearly 2450 performance against a field of masters. Both players will gain about 40 USCF rating points. The history of both over the past year clearly shows them as rapidly improving. Even with those 40 points, they are still very much underrated.
However, it is difficult to mathematically differentiate between a player who gains points in one event and then loses them all in the next two. Master class ratings have always been intended to be more steady than D player ratings. Wild swings in both directions are undesirable. In other words, increasing the K factor or number of bonus points for one tournament is not the solution.
Instead, the rating system could somehow incorporate a history over several tournaments, maybe the last 20 to 30 rated games. For each tournament, calculate a performance rating. If a player has a performance rating that consistently exceeds his actual rating by hundreds of points in these events, then the rating system can implement a super bonus. This super bonus would be a function of the player’s rating, meaning the super bonus becomes a factor only for high rated players (K factor under 20).
Interesting idea.
At first I thought you were suggesting an interpolation and adjustment based on strict chronological trajectory, almost a slope or acceleration calculation. But I think your suggestion is a little different.
.
Robert Hess has played in 25 events since 1/1/2006, during which time his rating has moved from around 2400 to 2531 after Foxwoods.
He lost rating points in 8 of those events. His rating went up 50 points (to 2531) based on his performance at Foxwoods, including 10 bonus points.
During that time period, his record against players rated below 2100 is 23-0-2,
his record against 2100 players is 11-1-3, his record against 2200 players is 20-1-9, his record against 2300 players is 15-1-5, his record against 2400 players is 8-7-8,
his record against 2500 players is 2-6-9, his record against 2600 players is 4-9-10
and his record against 2700 players is 0-5-3.
It seems to me like he has been performing in the 2400 range, as his performance against other 2400 players has been only slightly above .500 and he has a minus score against players rated 2500 or higher.
Sorry, I still don’t see any evidence that the ratings system isn’t working reasonably.
The Reno event doesn’t appear to be in yet, but since December of 2006 when he hit 2100, Daniel Naroditsky’s record by 100 point intervals is as follows:
Your contention that he has been playing at much over the mid to high 2200 level doesn’t appear to be consistent with his record against other masters, and prior to Reno his peak rating has been 2270.
You’re analyzing the entire 2 year period as if it were one chunk of time. Two years is a very long time in the development of a young player.
Analyze each six month period (or tournament by tournament) during those two years and let’s see if his rating at the beginning of the next six month period (tournament) was close to how he performed during that period. Sure, it can reflect accurately with enough games over years, but is it moving along fast, so that when one sits down to a game with him you know that rating on the wall-chart is fairly close (within 50-100 pts.)?
Mark, I really think you’ve formed a conclusion that cannot be supported by the facts.
Robert Hess played 66 games rated during 2007. Here’s a breakdown of his performance on 50 point intervals based on pre-event ratings for each of those events. (A negative number means he was rated below his opponent.)
Note that he had ZERO wins against players rated 100 or more points above him, regardless of what his rating was at the time.
